1. Limited time only! Sign up for a free 30min personal tutor trial with Chegg Tutors
    Dismiss Notice
Dismiss Notice
Join Physics Forums Today!
The friendliest, high quality science and math community on the planet! Everyone who loves science is here!

Job hunting advice for theoretical physics PhD

  1. Mar 21, 2013 #1
    I've often heard on PhysicsForums that career prospects are bright for theorists who paid attention to their math fundamentals and learned to program. That's me! It's also several of my fellow grad students. I suspect many others on this forum are in similar situations, and they may also benefit from advice.

    Nobody is calling us. So if employers aren't finding us, we'll look for them. The question is:

    What are some good ways to get our resumes, publications, portfolios, and/or selves in front of people who might hire us?

    For the sake of generality, please assume the following conditions. These are all true for me, and most are true for my colleagues.
    • We recently defended our PhDs or will do so within a few months.
    • Academic, government, and private jobs are all open for consideration. We are flexible about location. (Exception: No working for hostile governments. We will not help North Korea decrypt your email.)
    • We are competent programmers but lack the experience required for senior-level software jobs.
    • We do not care if the work is "boring." Blue-sky basic research is awesome, but plain old statistics is also perfectly good.
    • U.S. News did not rank our department "top tier," so we are not eligible for positions requiring degrees from high-status universities.
    • We have applied to postdocs but cannot afford to bet all-in on the classic academic postdoc/professor/tenure trajectory.
    • We do not live with our parents and therefore cannot afford to be adjunct instructors forever. (Adjuncts: Forgive my snark, but I think you know what I mean.)
    I've partitioned job-search methods into subsets. Pardon the violent analogies; we do not intend to actually shoot anyone. If I forgot anything, please say so.
    1. Buddy method: Ask friends/colleagues/family to introduce us to potential employers. Alas, this method is limited by the people we happen to know already.
    2. Sniper method: Email people who are experts at stuff we like. I've heard that it's better to ask for advice about a specific topic rather than say "hey, gimme a job."
    3. Shotgun method: Search for job postings and send a huge number of resumes and/or cover letters.
    4. Headhunter method: Email recruiters who specialize in industries related to our research and training.
    5. EDIT: Conference method: Go to professional and academic conferences. Talk shop with people and hand out business cards.
    I have used 1,2, and 3 with no results so far. I did get a few polite "Thank you, but your background is not an ideal fit for us" emails. This is odd because what those companies do is damn near isomorphic to what I do. But that phrasing may be a euphemism for "Lawyers advise us not to say exactly why we rejected you." I have not used #4 because I don't have reliable info about which recruiters are shady and which are mostly honest. If you know any good ones, please promote them here!

    In my case, I specialize in linear algebra, applied statistics, ordinary and stochastic differential equations, and a little cryptography and information theory. I do numerical simulations of all these things with NumPy, MATLAB, and Mathematica. I know more than enough C++ to code FizzBuzz, but not enough to impress a professional. I've been aiming for financial modeling, data mining, or quantum computing - but other suggestions are welcome.

    My colleagues' abilities differ, but they're all good at basic probability, numerical programming, and either PDEs or ODEs. Most of them are also good teachers and skilled at explaining complicated things simply without dumbing them down or insulting the audience.

    Any practical advice is greatly appreciated! (Ideological nonsense will be quietly ignored.)
    Last edited: Mar 21, 2013
  2. jcsd
  3. Mar 21, 2013 #2

    Vanadium 50

    User Avatar
    Staff Emeritus
    Science Advisor
    Education Advisor
    2017 Award

    Let's start with the simple questions. If Company XXX hires you, how exactly is that going to help their bottom line?
  4. Mar 21, 2013 #3


    User Avatar
    Science Advisor
    Education Advisor

    I'll just throw a couple ideas out there...

    1. Have you considered or attended professional conferences? I don't know who this would be for the markets you're talking about (South by Southwest maybe?), but if you wanted to get into, for example, medical imaging companies you may want to try attending RSNA's annual meeting, AAPM, or ASTRO. This unfortunately requires travel and conference registration, but in my experience these are great opportunities for face to face networking.

    2. Job shadows? Rather than asking for an immediate position, you may want to just try to arrange some job shadowing days. This can allow you to meet people outside your immediate network and talk with people doing the work you're interested in. Even if that particular company is not hiring, people in the field are usually aware of who is hiring or at least what the market is like.
  5. Mar 21, 2013 #4
    I suppose that depends on the specific industry. For example, suppose XXX is a hedge fund with lots of time-series data on historical securities prices ##S_t##. I could write a program to find the securities whose ##\log(S_t)## has more large jumps than are predicted by a model (e.g. Black-Scholes, SABR) given its previous volatility. Then I could write a report warning traders that those securities are probably more risky than the models indicate.

    Or maybe XXXCorp makes widgets, and demand for widgets is known to be stronger in the winter. I could deconvolve the widget demand history to produce a temperature-adjusted plot of widget sales so the company can better estimate how much of their recent gains/losses are just natural seasonal variation.

    If teamed with a database expert (who need not be a statistician), I could do either of these things quickly for a large variety of different securities/widgets.
  6. Mar 21, 2013 #5
    I was going to include that as "Conference method," but I totally forgot. I'm hoping to do that soon - though as you mentioned, it's tough when you're short on time and money.

    That sounds like a very good idea. Do you know any good methods for arranging a job-shadow? My first guess is to cold-email companies, but I'm not sure who exactly to email or what the etiquette is.
  7. Mar 21, 2013 #6

    Vanadium 50

    User Avatar
    Staff Emeritus
    Science Advisor
    Education Advisor
    2017 Award

    OK, now what you need to do is focus on what you can do for the company - how are you going to make them money. That;'s what they are interested in.
  8. Mar 22, 2013 #7
    So I tried variants of all of the above with fairly limited success. Eventually I was given the following advice, which did work
    1. If you have done work that isn't directly industry relevant, give up on engineering/physics related work- there are far too many "ready-hires" with directly relevant backgrounds and you can't beat them out. (maybe the small improvement in the economy have made this easier).

    2. find an in-demand field not opposed to physicists and sit down and learn the relevant field, and then start applying.

    So while I was working as a bartender, I worked through some statistics and data-mining relevant stuff until I felt I was competent and then started applying. In the end, though, a fortuitous bit of luck landed me my first job- I was talking with a customer, it turned out he was looking to hire for a data mining group at an insurance company. He gave me his card, and I called the next day and was hired.

    So if you want to do finance work, work through a few derivatives textbooks, code up a small project of your own, and then start applying. If you want to do datamining/stats work, learn the methods and again, code up a small project of your own. Basically, realize no one will hire you because you know physics, so you better learn something useful.
  9. Mar 22, 2013 #8
    I thought exactly the same thing, so I did a few mini-projects. One of them is even datamining/stats, though I suspect it will fail to impress a data science professional. I designed two statistical tests of online poker shuffles, and an astro PhD candidate extracted the relevant data from my Full Tilt hand history files. It's here if you're curious.

    I think I'll try to combine your advice with Choppy's. The numerical method in my thesis can be used for commercial applications like autopilot error-correction and pricing rainbow options. I want to release it open-source, but I don't know how to advertise it to people who will actually use it. If I can afford it, I might be able to present it at engineering and/or finance conferences.
    That's kind of terrifying: you did all of these methods, and none of them worked as well as bartending!

    This is way off-topic, but I can't resist... when you make an Old-Fashioned for yourself, do you mash fruit in the bottom, or just toss in a lemon/orange peel? (I have a friend who's a neutrino hunter and her husband is a bartender. They skip the fruit, but I go either way.)
  10. Mar 22, 2013 #9


    User Avatar
    Science Advisor
    Education Advisor

    As far as the job-shadowing thing goes, you might want to ask at your school's student placement services office. Arranging job shadows is generally one of their functions so they'll know who to talk to. Personally I would avoid trying to go through HR and try to see if you can email someone as close to the position that you're actually interested in getting. I think cold emailing is fine (I get them every once in a while), and I don't really know too many alternatives.

    One rule of thumb I've discovered is that people like to talk about themselves. So even if a full day job shadow is out of the question, most people would be willing to meet for an hour or so over lunch or coffeee to talk about why they do and their experiences in the field.
  11. Mar 22, 2013 #10
    Always a lemon peel, never mashed fruit. Are you mashing fruit in place of the sugar? But hardly anyone ever orders an Old-Fashioned, and my preference for drinking is just whiskey on the rocks. I think I made less than 20 in several years of bartending.

    For the sort of work I do, I'd suggest grabbing the data set for one of the competitions on kaggle.com and putting together a predictive model (a neural net, a random forest,etc) of some kind. I think to land a job you need to be able to answer some general questions about the process of building a model (how many features would you use in a model with X data points?, if you had hundreds of potential features, how would you narrow them down? what sort of model/approach would you try and why? Where would you expect your model to break down?) If you've worked through even one model, you probably have more context to answer these.
  12. Mar 22, 2013 #11
    It always confuses me when people talk about programming as if it's some binary thing where you either know it or you don't. It's a huge range of skill levels. If you're highly skilled at it- and you can prove it- then you'll have no problem finding programming jobs. If you only know the very basics (ie "hello world", arithmetic loops, that kind of thing) then that's not useful at all. If you're in between then it just depends on what specifically you've learned and whatever companies are hiring for at the moment. But just saying "I know how to program" is like saying "I know physics- can I be a physicist?"
  13. Mar 22, 2013 #12
    Because if you're doing programming for theoretical physics applications, you are likely a). at least of intermediate skill, and b). may have even developed specifically useful skills. This is my experience working in a comp biophysics lab.
  14. Mar 22, 2013 #13
    Bit of trivia: Major Howard Armstrong, inventer of the Superheterodyne receiver and wideband FM drank an Old Fashioned with dinner almost every night. (he was the type of engineer to wear the same clothes and eat the same meals every day)
  15. Mar 22, 2013 #14
    What simulation code would be just a "for loop"?

    In a research setting you tend to need to make adjustments ( a lot of prototyping) to your model or speed up your implementation so that you would have to go out of your way to just learn basic programming in the 5+ years that a PhD program takes.
  16. Mar 23, 2013 #15


    User Avatar

    I work as a programmer and I can tell you that for generic business programming type of stuff, the market is heating up. The software field is faddish though so you have to have skills that are in demand now. Things like Python and Ruby are in now....a few years ago it was Java, .NET etc. But just knowing a programming language won't get you far...you need things like good OO methodology, design patterns...these things take time and are best studied within a project. If you have free time, consider actively contributing to an open source project.
    And learn SQL
  17. Mar 23, 2013 #16
    I'm going to visit the career center soon. I'll be sure to ask them about that sort of thing.
    That's definitely the plan. If I can talk to one of the lead math/stats/programming people, they'll get a lot more information than is contained on a resume. IMO for many subjects, most experts need < 15 minutes to approximately figure out whether someone actually knows the stuff they mentioned on their application.

    My cold-email method is to ask a short, legitimate question that requires the recipient's expertise but not much of their time. For example: Do you think rejection-sampling RNGs like Ziggurat are worth the extra coding work? I'm using Marsaglia-polar and it seems fine, but I thought I should ask a professional. It gives them an excuse to talk about themselves and/or their subject without seeming rude.

    For most of the competent TAs, emails like that from students are more likely to be answered. But there's a classic signal-to-noise problem with cold emails because they so often get ignored. When they do, it's hard to tell whether it's bad luck or I'm doing it wrong.
  18. Mar 23, 2013 #17
    More precisely, I meant "We are competent programmers [in some language(s)] but lack the experience required for senior-level software jobs."
    That's an accurate description of most of the physicists I know. Some of the biophysics people are very skilled at CUDA multithreading, and I've released a state-of-the-art open-source linear ODE solver. But few of us have ever made a working GUI.
    If I'm unemployed for a few months, I might use a convex combination of this advice and what ParticleGrl said. I know enough math and Python that I might be able to contribute something useful to NumPy/SciPy if I have the time.

    Do you have an idea how long it takes to get competent at using SQL? I was going to practice C++ and R instead. I already know some C++ and a ton of statistics, so I have a head-start on these languages.
  19. Mar 23, 2013 #18
    SQL is pretty trivial. I remember when I first started learning it I found the logic of nested queries a bit weird, but it wasn't difficult to get through. Even big table joins and such are pretty simple. Maybe there's some super-use for the language that I'm not familiar with, but I've been using it for over four years and I get the job done.

    Some of it will depend on how much work you're needing your queries to do for you. Personally, I prefer to manipulate data myself; I bring back as much data as I can handle and that doesn't lend itself to complicated queries.
  20. Mar 23, 2013 #19
    So in the sort of work I do, 'as much as I can handle' (in memory) is usually significantly smaller than the whole data set, so I have to do the heavy lifting in SQL. Even so, I came in with minimal experience and have never had any trouble accomplishing what I want. Occasionally, I probably have to take a few more looks at optimization than more experienced people, but I at least get to the finish.
  21. Mar 23, 2013 #20


    User Avatar

    You don't need to. With your skill set and training, you would be a more natural fit for server side or "backend" programming
    Depends by what you mean by "competent". It will take you probably a few days to learn the syntax for standard sql and be able to do simple queries. But it can take years to be an expert database developer or dba. For most developers, though, I'd say you can be reasonably expert after six months of use.
  22. Mar 26, 2013 #21
    I started doing a mini-project which is kind of like that, I think. It's basically a simulated general practitioner: it asks a bunch of questions, looks at a medical database, and uses Bayesian inference to guess what the problem is. It's not on Kaggle, but I pitched it to a medical start-up who might actually use it or something based on it.

    I'll keep an eye on Kaggle, though. If I can find the time, I might even be able to help with their SciKit-Learn project because I've already done a big pile of linear algebra in NumPy.

    I just got one of those Teach Yourself SQL In 1 Femtosecond books. Based on my tiny experience so far, it seems consistent with what Locrian and jk said: a few days to do basic stuff, and a long time to be a Grandmaster.

    I'm hoping to find work that's heavy on numerical algorithms and light on advanced comp-sci, so I might be OK just knowing basic relational-database stuff for now. When one has a titanium screwdriver, it's tempting to solve every problem with a screwdriver and neglect all the other tools.

    I stir a small teaspoon of brown sugar, Angostura, and Peychaud's in the glass with a little splash of water. Then I rip a real (not maraschino) cherry in half, toss out the pit, toss in an orange slice, and stab at the fruit with a spoon because I don't have a muddler. Or when I don't have any cherries, I do what you said. I think that's the old-fashioned Old-Fashioned.
  23. Mar 26, 2013 #22
    Someone off-topic but related question: Does anyone feel that part of the problem physics Ph.D's have in finding employment in industry outside of their field of expertise is that a lot of hiring managers feel threatened by Ph.D. level physicists? Not necessarily by their skill set alone, but more what the ability to acquire that skillset implies about their talent and intelligence.
  24. Mar 26, 2013 #23
    That should work. Just be sure you can answer questions like 'why did you pick Bayesian inference over other models/approaches', and make sure you held-out some of the rows from the database and used them to validate so you could create some metrics on how well it performs.

    If you are going to the trouble of making cocktails, you should get a muddler, a cheap one is <$5, and you will be the envy of your friends.

    Also, unlike the link (weird to see Rachel Maddow in that context), I generally wouldn't use any soda water, just a sugar cube and some bitters muddled together, then whiskey then a bit of lemon rind.
    Last edited: Mar 26, 2013
  25. Mar 26, 2013 #24


    User Avatar


    My take is that the difficulty that PhD's have in finding jobs in industry has more to do with their lack of skills that are immediately useful to hiring managers.

    Incidentally, although I don't have a PhD, I believe the acquisition of a PhD in any field says more about a person's capacity for hard work and perseverance than their talent or intelligence, however vague those terms are.
  26. Mar 26, 2013 #25

    The real problem is coming in with this attitude when, in reality, you are probably underqualified for the position.
Share this great discussion with others via Reddit, Google+, Twitter, or Facebook